Ap4 is rate limiting for intestinal tumor formation by controlling the homeostasis of intestinal stem cells

(1)

Ap4 is rate limiting for intestinal tumor formation

by controlling the homeostasis of intestinal stem

cells

Stephanie Jaeckel

1 , Markus Kaller

1 , Rene Jackstadt

1 , Ursula Götz

1 , Susanna Müller

2 , Sophie Boos

3,4,5

,

David Horst

2,4,5,6

, Peter Jung

3,4,5

& Heiko Hermeking

1,4,5

The gene encoding the transcription factor TFAP4/AP4 represents a direct target of the

c-MYC oncoprotein. Here, we deleted Ap4 in Apc

Min

mice, a preclinical model of inherited

colorectal cancer. Ap4 de

ﬁciency extends their average survival by 110 days and decreases

the formation of intestinal adenomas and tumor-derived organoids. The effects of Ap4

deletion are presumably due to the reduced number of functional intestinal stem cells (ISCs)

amenable to adenoma-initiating mutational events. Deletion of Ap4 also decreases the

number of colonic stem cells and increases the number of Paneth cells. Expression pro

ﬁling

revealed that ISC signatures, as well as the Wnt/

β-catenin and Notch signaling pathways are

downregulated in Ap4-de

ﬁcient adenomas and intestinal organoids. AP4-associated

sig-natures are conserved between murine adenomas and human colorectal cancer samples. Our

results establish Ap4 as rate-limiting mediator of adenoma initiation, as well as regulator of

intestinal and colonic stem cell and Paneth cell homeostasis.

DOI: 10.1038/s41467-018-06001-x

OPEN

1_{Experimental and Molecular Pathology, Institute of Pathology, Ludwig-Maximilians-Universität München, Thalkirchner Strasse 36, D-80337 Munich,}

Germany.2Institute of Pathology, Ludwig-Maximilians-Universität München, Thalkirchner Strasse 36, D-80337 Munich, Germany.3DKTK Research Group, Oncogenic Signaling Pathways of Colorectal and Pancreatic Cancer, Institute of Pathology, Ludwig-Maximilians-Universität München, Thalkirchner Strasse 36, D-80337 Munich, Germany.4_{German Cancer Consortium (DKTK), Partner site Munich, Munich D-80336, Germany.}5_{German Cancer Research Center}

(DKFZ), Heidelberg D-69120, Germany.6_{Institute of Pathology, Charité}_{– Universitätsmedizin Berlin, Charitéplatz 1, D-10117 Berlin, Germany.}

Correspondence and requests for materials should be addressed to H.H. (email:heiko.hermeking@med.uni-muenchen.de)

123456789

(2)

T

he TFAP4/AP4 protein belongs to the class of

basic-helix-loop-helix leucine zipper (bHLH-LZ) transcription factors

(reviewed in Jung and Hermeking

1

_{). AP4 exclusively forms}

homodimers, which bind to the E-box motif CAG/CCTG and

thereby either repress or activate the expression of target genes.

We previously identiﬁed the AP4 gene as a direct transcriptional

target of c-MYC

2

. AP4 is expressed in

progenitor/transient-amplifying (TA) cells in human colonic crypts, and in colorectal

cancer (CRC) in a pattern similar to c-MYC. The prototypic

oncogene c-MYC is a direct target of the APC (adenomatous

polyposis coli) /Wnt (Wingless/Int-1) pathway

3

and an essential

mediator of tumor formation induced by inactivation of Apc in

the intestine

4,5

_{. Previous studies performed in CRC cell lines or}

mouse embryonic

ﬁbroblasts suggested that AP4 may contribute

to the progression of CRC by regulating genes involved in

epithelial–mesenchymal transition (EMT) and proliferation

6–8

_.

However, the organismal function of Ap4 in the intestinal

epi-thelium and its relevance for intestinal tumor formation has so far

not been studied using a genetic approach.

The present study shows that inactivation of Ap4 by deletion

leads to decreased adenoma formation in Apc

Min

mice, which

represent a preclinical model of familial adenomatous polyposis

(FAP)

9,10

. mRNA proﬁling revealed downregulation of a large

number of genes involved in Wnt/β-catenin and/or Notch

sig-naling in Ap4-deﬁcient adenomas of Apc

Min

_{mice and organoids}

derived from the epithelium of the small intestine. In line with

these regulations, Ap4-deﬁcient intestinal organoids and

tumor-oids show impaired re-growth capacities and therefore decreased

stemness. The reduced number of tumors observed in

Ap4-deﬁ-cient mice is presumably due to a decrease in the number of

functional intestinal stem cells (ISCs). In addition, Ap4

inacti-vation causes an increase in the number of Paneth cells. Our

results establish Ap4 as a regulator of ISC and Paneth cell

homeostasis and as a rate-limiting mediator of intestinal tumor

initiation.

Results

Role of

Ap4 in intestinal adenoma formation. Here we

deter-mined the effect of Ap4 deﬁciency on adenoma formation in the

intestine of Apc

Min

_{mice, which harbor an inactivating mutation}

in one Apc allele. Upon spontaneous loss of the second Apc allele,

these mice develop ~50–100 adenomas in the small intestine by

the age of 4–6 months. As expected, adenomas in Apc

Min/+

_/Ap

-/-mice did not display Ap4 expression, whereas adenomas of

Apc

Min

mice showed elevated expression of Ap4 (Fig.

1 a).

Approximately 50% of Apc

Min

mice succumbed to intestinal

adenomas by ~180 days of age (Fig.

1 b), which was in line with

previous reports

11,12

. However, in Ap4-deﬁcient Apc

Min

mice

intestinal cancer-related death was delayed on average by

110 days, with heterozygous mice showing an intermediate delay.

Ap4 deﬁciency was associated with a ~4-fold decrease in the

number of adenomas in the small intestines of moribund Apc

Min

mice, while the size of adenomas increased (Fig.

1 c–e).

Unex-pectedly, the proliferation rate within small intestinal adenomas

of moribund Apc

Min

mice was not affected by loss of Ap4

(Sup-plementary Fig. 1). When adenomas of age-matched, 120 days old

Apc

Min

_{mice were compared, the Ap4-deﬁcient mice showed a}

~5-fold decrease in the number of adenomas, whereas the size of

the adenomas was not affected (Fig.

2 a-c). A decreased number of

tumors was also detected in the colon of Ap4-deﬁcient Apc

Min

mice when compared with Ap4-wild-type Apc

Min

mice

(Supple-mentary Fig. 2a). However, due to the low incidence of adenomas

in the colon of Apc

Min

_{mice these differences did not reach}

sta-tistical signiﬁcance. The uniform tumor size and the unchanged

proliferation rate of tumors in the small intestine (Supplementary

Fig. 2b) in 120 days old mice suggested that the increase in

adenoma size seen in moribund animals was most likely due to

the increased life-span of Ap4-deﬁcient Apc

Min

_{mice. The effects}

of Ap4 loss on tumorigenesis observed in Apc

Min

mice were

independent of the gender (Supplementary Fig. 2c-e). When we

analyzed Apc

Min

_{mice with intestinal epithelial cell (IEC)-speciﬁc}

deletion of Ap4, which was achieved by crossing Villin-Cre with

Ap4

ﬂ/ﬂ

mice, we obtained similar results as for Apc

Min

mice with

germ-line deletion of Ap4: that is, in Ap4

ΔIEC

/Apc

Min

_mice

intestinal cancer-related death was signiﬁcantly delayed by

110 days, with heterozygous mice showing an intermediate delay

(Supplementary Fig. 2f). Ap4

ΔIEC

/Apc

Min

_{mice showed a sixfold}

decrease in the number of adenomas in the small intestines of

moribund and a

ﬁvefold decrease in the small intestine of

120 days old Apc

Min

mice, whereas the size of adenomas increases

in moribund mice and the size of the adenomas was not affected

in 120 days old mice (Supplementary Fig. 2g, h).

Epithelial-speciﬁc deletion of Ap4 in Apc

Min

_{mice also resulted in a}

decreased number of adenomas in the colon, although this effect

was not statistically signiﬁcant (Supplementary Fig. 2i). Deletion

of Ap4 in epithelial cells did not affect the proliferation rate of

established adenomas of Apc

Min

_{mice (Supplementary Fig. 2j).}

Therefore, the effects of Ap4 deletion in the germ-line on

ade-noma formation are presumably intestinal epithelial cell

auton-omous. Taken together, these results show that Ap4 is rate

limiting for adenoma initiation in Apc

Min

_{mice. As c-Myc is a}

required mediator of intestinal tumor formation in Apc-mediated

tumorigenesis, the results imply a pivotal role of Ap4 among the

many known c-Myc target genes in mediating intestinal tumor

formation.

mRNA expression pro

ﬁling of Ap4-deﬁcient adenomas. To

identify pathways mediating the effects of Ap4, we compared the

mRNA expression proﬁles of intestinal adenomas that formed in

Apc

Min

mice with and without Ap4 deletion. By applying

next-generation sequencing (NGS), we identiﬁed 1459 mRNAs that

were differentially regulated with a fold change in expression > 1.5

(p < 0.05) due to deletion of Ap4 (Fig.

3 a). Out of these, 954

mRNAs were signiﬁcantly downregulated, and 505 mRNAs were

upregulated in Ap4-deﬁcient Apc

Min

_{mice (Fig.}

₂

_{b, c). Notably,}

pathway analysis showed that mRNAs encoding for proteins

involved in EMT, as well as cell cycle regulatory proteins (e.g.,

E2F targets) were signiﬁcantly enriched among the

down-regulated mRNAs (Supplementary Fig. 3a, Supplementary

Data 1). Furthermore, gene set enrichment analysis (GSEA)

showed that mRNAs characteristic for Lgr5-positive ISCs

13

were

preferentially downregulated in Ap4-deﬁcient adenomas (Fig.

4 a,

b). We validated this

ﬁnding using additional, previously

pub-lished ISC-speciﬁc gene signatures

14,15

_{(Supplementary Fig. 3c,}

Supplementary Data 2). These signatures also showed preferential

enrichment among the mRNAs downregulated after deletion of

Ap4. Moreover, mRNAs encoding for proteins involved in Wnt/

β-catenin and Notch signaling, which control the homeostasis of

ISCs

16,17

_{, were also preferentially downregulated in Ap4-deﬁcient}

adenomas (Fig.

4 a, b, Supplementary Fig. 3c, Supplementary

Data 2). Genes downregulated upon deletion of Ap4 included ISC

markers induced by Wnt/β-catenin signaling, such as Lgr5 and

Ascl2

18–20

, or by Notch signaling, such as Olfm4

21

, as well as

additional direct Wnt/β-catenin and/or Notch target genes with

critical functions in the Wnt and Notch signaling pathways, such

as Sox4, Tcf7/Tcf1, Axin2, EphB3, Jag1, Jag2, Hes1 and c-Myc

(Fig.

4 b). Furthermore, Notch1 itself was downregulated in

Ap4-deﬁcient adenomas. Taken together, these results imply that Ap4

regulates the homeostasis of ISCs via activating Wnt/β-catenin

and/or Notch signaling pathways.

(3)

Overall survival (%) Days 0 100 200 300 400 500 600 0 20 40 60 80 100 Tumor diameter (mm) 0 2 4 6 150 100 50 0 Duodenum Jejunum

IIeum Colon Sum.

Duodenum Jejunum

IIeum Colon Sum.

Moribund Moribund Adenomas/mouse n.s. n.s. n.s. n.s.

***

a

b

Ap4 Ap4

c

d

e

*

* *

***

**

*

**

***

* *

***

ApcMin/+_/Ap4+/+ _{(n = 79)}

ApcMin/+/Ap4+/+

ApcMin/+_/Ap4+/+ _ApcMin/+_/Ap4+/ – _ApcMin/+_/Ap4–/ – _ApcMin/+_/Ap4+/+ _ApcMin/+_/Ap4+/ – _ApcMin/+_/Ap4–/ –

ApcMin/+_/Ap4+/+

ApcMin/+_/Ap4+/ –

ApcMin/+_/Ap4–/ –

ApcMin/+_/Ap4+/– _{(n = 79)}

ApcMin/+_/Ap4–/– _{(n = 58)}

Fig. 1 Deletion of Ap4 in ApcMin/+mice prolongs survival and decreases the frequency of adenomas.a Immunohistochemical detection of Ap4 in adenomas of moribund ApcMin/+mice with the indicated genotype. Counterstaining with hematoxylin. Scale bar= 50 µm. b Kaplan–Meier survival analysis of ApcMin/+mice with the indicated genotypes. Censored mice without intestinal tumor-related death are indicated on the Kaplan–Meier curve as tick marks.c Macroscopic pathology of representative polyps in the small intestine (ileum) of moribund ApcMin/+mice with the indicated genotype, scale in cm.d Representative sections through rolls of the small intestine stained for hematoxylin and eosin (HE). Scale bar= 500 µm. e Quantification of adenoma number/mouse (left panel) and tumor diameter (right panel) in the intestine of six male and six female (ApcMin/+/Ap4+/+),five male and five female

(ApcMin/+/Ap4+/-) or four male and four female (ApcMin/+/Ap4-/-_{) moribund Apc}Min/+_{mice. The box plot extends from the 25th to 75th percentiles. The}

line in the middle of the box is plotted at the median. The whiskers underneath or above the boxes range from min. to max. value, respectively.b Results were subjected to a log-rank test with p-values * < 0.05, ** < 0.01, *** < 0.001, n.s. not signiﬁcant. e Results represent the mean ± SD. Results were subjected to an unpaired, two-tailed Student_{’s t-test with p-values * < 0.05, ** < 0.01, *** < 0.001, n.s. not signiﬁcant. See also Supplementary Fig. 1}

(4)

Recently, Ap4 was shown to maintain a c-Myc-induced

transcriptional program in activated T cells

22

_{and germinal}

center B cells

23

. In line with these

ﬁndings, c-Myc target genes

were preferentially downregulated in Ap4-deﬁcient adenomas

(Fig.

4 a; Supplementary Data 2). However, the changes in

expression of c-Myc target genes observed after deletion of Ap4

were rather modest compared to the regulations observed in ISC

signature or Notch signaling components (Supplementary

Fig. 3d). Similarly, E2F targets, though signiﬁcantly enriched

among the downregulated RNAs (Supplementary Fig. 3c),

displayed only modest changes in expression that were

compar-able to those of c-Myc targets (Supplementary Fig. 3d). These

modest regulations of c-Myc and E2F targets may explain the

lacking inﬂuence of Ap4 deletion on cell proliferation within

adenomas.

We exemplarily conﬁrmed the differential regulation detected

by NGS using quantitative PCR (qPCR). Thereby, we validated

the downregulation of the stem cell markers Smoc2, Lgr5 and

Olfm4, as well as the repression of several genes involved in the

Wnt/β-catenin signaling and/or Notch signaling in Ap4-deﬁcient

adenomas (Fig.

4 c). Consistent with its previously reported

repression by AP4

2

, Cdkn1a/p21 was upregulated in

Ap4-deﬁcient adenomas. Interestingly, we did not detect a change in

mRNA or protein levels of Ctnnb1 (β-catenin) in Apc

Min

adenomas (Fig.

4 c, Supplementary Fig. 3e), suggesting that Ap4

directly regulates Wnt/β-catenin target genes.

Next, we analyzed whether Ap4 directly regulates the

expression of ISC markers and components of the Wnt/β-catenin

and/or Notch signaling pathways. Our analysis of Ap4

DNA-binding patterns in murine T and B cells

22,23

_{revealed Ap4}

occupancy within the promoter regions of Ascl2, Axin2, c-Myc,

Dll1, Dll4, EphB3, Hes1, Hey1, Jag1, Jag2, Notch1, Sox4, and Tcf7

(Supplementary Fig. 3f). We performed quantitative

chromatin-immunoprecipitation (qChIP) analysis to conﬁrm Ap4

occu-pancy in the murine CRC cell line CT26 at the promoters of the

following genes: Ascl2, Dll1, Dll4, EphB3, Hes1, Jag1, Jag2, Notch1,

Sox4 and Tcf7 (Fig.

4 d). Similar to the promoter of human

CDKN1A/P21, the murine Cdkn1a/p21 promoter also contains

Ap4-binding sites that showed occupancy by Ap4 (Fig.

4 d).

Therefore, Cdkn1a/p21 is a conserved, direct Ap4 target. Taken

together, these results suggest that the differential regulation of

genes involved in Wnt/β-catenin and/or Notch signaling

observed in Ap4-deﬁcient Apc

Min

_{adenomas is a direct}

con-sequence of the absence of Ap4 at the respective promoters.

Analysis of tumor organoids from

Ap4-deﬁcient Apc

Min

mice.

Our NGS results suggested that Ap4 is involved in maintaining a

stem cell-like expression pattern in tumor cells. This was

con-ﬁrmed by in situ hybridization with a probe detecting the mRNA

expression of Lgr5 (Fig.

5 a) or Smoc2 (Supplementary Fig. 4a) in

small intestinal adenomas of Apc

Min

mice. Indeed, Ap4-deﬁcient

adenomas displayed less cells positive for Lgr5 or Smoc2 mRNA

expression when compared with adenomas expressing Ap4.

Therefore, the number of tumor stem cells is presumably

decreased in the absence of Ap4. Next, we generated tumor

Tumor diameter (mm) 0 2 4 6 Adenomas/mouse 100 150 50 0 120 days

*

120 days n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. _n.s. n.s. n.s. n.s. n.s.

a

b

c

* *

***

**

***

Duodenum Jejunum

IIeum Colon Sum.

Duodenum Jejunum

IIeum Colon Sum.

ApcMin/+_/Ap4+/+ _ApcMin/+_/Ap4+/ – _ApcMin/+_/Ap4–/ – _ApcMin/+_/Ap4+/+ _ApcMin/+_/Ap4+/ – _ApcMin/+_/Ap4–/ –

ApcMin/+/Ap4+/+ ApcMin/+_/Ap4+/ –

ApcMin/+/Ap4–/ –

Fig. 2 Ap4 deletion decreases frequency but not size of adenomas in age-matched ApcMin/+mice.a Macroscopic pathology of representative polyps in the small intestine (ileum) of 120 days old ApcMin/+mice with the indicated genotype, scale in cm.b Representative sections through rolls of the small intestine were stained forβ-catenin. Scale bar = 500 µm (upper pictures) or 250 µm (lower pictures). c Quantiﬁcation of adenoma number/mouse (left panel) and tumor diameter (right panel) in the intestine of four male and four female 120 days old ApcMin/+mice per genotype. The box extends from the 25th to 75th percentiles. The line in the middle of the box is plotted at the median. The whiskers represent the minimal and maximal values.c Results represent the mean ± SD. Results were subjected to an unpaired, two-tailed Student’s t-test with p-values * < 0.05, ** < 0.01, *** < 0.001, n.s.: not signi_{ﬁcant. See also Supplementary Fig. 2}

(5)

organoids using single cells derived from adenomas of Apc

Min

mice. Indeed, cells directly isolated from Ap4-deﬁcient adenomas

formed tumor organoids ex vivo with a signiﬁcantly lower

fre-quency than those derived from Ap4 wild-type adenomas

(Fig.

5 b). However, after the

ﬁrst passage ex vivo the

Ap4-defi-cient and -profiAp4-defi-cient tumoroids re-built new tumor organoids

with a comparable frequency and growth rate (Fig.

5 b).

There-fore, Ap4 appears to be required for the initiation, but not for the

maintenance of ex vivo cultured intestinal tumoroids. The

dele-tion of Ap4 in these tumor organoids was accompanied by lower

levels of the ISC markers Smoc2, Lgr5 and Olfm4 when compared

with Ap4 wild-type adenomas (Fig.

5 c). Additional genes involved

in Wnt/β-catenin signaling and/or Notch signaling, including

Notch1 itself, were also downregulated in Ap4-deﬁcient tumor

organoids (Fig.

5 d). Also at the protein levels, the cleaved, active

form of Notch1 (NICD1) and Hes1, which is encoded by a Notch

target gene, were decreased in Ap4-deﬁcient tumor organoids

indicating a decrease in Notch signaling (Fig.

5 e). Therefore, the

reduced de novo tumor organoid formation capacity may be

caused by the downregulation of genes required for in vivo ISC

function upon Ap4 loss.

Subsequently, we isolated small intestinal crypts from

Lgr5-Cre

ERT2

_/Apc

ﬂ/ﬂ

_{and Lgr5-Cre}

ERT2

_/Apc

ﬂ/ﬂ

_/Ap4

ﬂ/ﬂ

_{mice. After}

plating of crypts in Matrigel overlaid with ENR media (containing

epidermal growth factor (EGF), Noggin and RSPO1), we acutely

deleted Apc or Apc in combination with Ap4 in Lgr5-positive

stem cells of newly formed intestinal organoids by addition of

4-hydroxy-tamoxifen (4-OHT). After passaging (passage 1) and

seeding the same amount of cells per drop of Matrigel, we

switched culture conditions to EN media devoid of RSPO1, in

which only Apc-deﬁcient tumoroids can grow. We obtained less

de novo formed tumoroids after Ap4 deletion when compared

with Ap4-proﬁcient tumoroids (Fig.

5 f, g). This supports the

notion that Ap4 has an important role during tumor initiation

and conﬁrms the result we previously obtained in vivo.

Reassur-ingly, tumoroids did not form in the absence of RSPO1 and

4-OHT (Supplementary Fig. 4b). During serial passaging, the

amount and size of tumoroids was not inﬂuenced by the deletion

of Ap4 (Fig.

5 f, g, h and Supplementary Fig. 4c). To exclude the

possibility that Ap4-deﬁcient tumoroids grew due to incomplete

deletion of Ap4, the complete deletion of Ap4 was conﬁrmed by

genomic PCR (Supplementary Fig. 4d). Taken together, these

results conﬁrm a critical role of Ap4 in the initiation, but not for

the maintenance of the tumoroids. These results are in line with

the observations initially obtained with Apc

Min

mice, where Ap4

loss decreased the number of adenomas but not their size.

Ap4 regulates the homeostasis of ISCs. Next, we analyzed the

expression of Ap4 and the effect of Ap4 deletion in the normal,

murine intestine. Ap4 protein was detected at the crypt base and

in TA cells located above the crypt base in the small intestine

(Fig.

6 a, left panel). As expected, Ap4 expression was not

detectable in the intestinal epithelia of Ap4 knock-out mice,

a

1 2 3 1 2 3 Min Max Rel. expression n = 954 n = 505 n = 505 n = 954

ApcMin/+_/Ap4fl/fl _ApcMin/+_/Ap4ΔIEC

ApcMin/+_/Ap4ΔIEC _{vs. Apc}Min/+_/Ap4fl/fl

1459 mRNAs DESeq2 (1769 mRNAs) edgeR (2090 mRNAs) Fold change ≥1.5× p < 0.05 –log 10 (p -value) 0 2 6 –4 –2 0 2 4 8

log₂fold change mRNAs

4

b

c

Fig. 3 Expression analyses of AP4-deficient adenomas from ApcMin/+mice.a Venn diagram displaying differentially regulated RNAs (fold change⩾ 1.5, p < 0.05) in Ap4fl/fland Ap4ΔIECApcMin/+adenomas as determined by edgeR and DESeq2.b Volcano plot and heatmap depicting expression changes between ApcMin/+/Ap4fl/fland ApcMin/+/Ap4ΔIECtumors from 120 days old mice derived from three female mice (five tumors per mouse) per genotype detected by RNA-Seq. Volcano plot: p-values are plotted against the log2of the corresponding RNA expression changes in Ap4ΔIECversus Ap4fl/fl adenomas. Differentially expressed RNAs (p-value < 0.05) with a log2fold change≥ 0.58 are indicated in red, with a log2fold change≤ −0.58 are marked in blue. RNAs with 0.58 > log2fold change >−0.58 and/or with a p-value ≥ 0.05 are represented by gray dots. Dashed vertical lines indicate cut-offs for differential expression. Dashed horizontal line indicates the cut-off for adjusted p-values < 0.05 as determined with DESeq2.c Heatmap depicting expression changes of differentially expressed mRNAs (fold change⩾ 1.5 and p < 0.05 as determined by edgeR and DESeq2) as relative expression levels normalized to the mean expression in the control, ApcMin/+/Ap4fl/fl, samples for each indicated mRNA. Colors indicate relative expression values from minimum (blue) to maximum (red) for each RNA sample per differentially regulated mRNA. Three biological replicates per genotype were analyzed

(6)

indicating that the antibody used here is speciﬁc for Ap4 (Fig.

6 a,

right panel). In mice expressing enhanced green

ﬂuorescent

protein (eGFP} from an Lgr5-promoter, Ap4 expression was

detected in eGFP-positive ISCs and TA cells, but not in the

adjacent lysozyme-positive Paneth cells (Supplementary Fig. 5a).

In Ap4-deﬁcient mice, the number of eGFP-positive ISCs was

signiﬁcantly decreased, indicating that Ap4 is necessary for ISC

maintenance (Fig.

6 b). Notably, Ap4-deﬁcient mice displayed a

signiﬁcant decrease in the number of ISCs positive for Olfm4

mRNA expression (Fig.

6 c). Furthermore, they showed an

increased number of Paneth cells in all regions of the small

intestine (Fig.

6 d): in the ileum each crypt section contained ~8

b

a

Stem cell signature Wnt/β-cat. components Notch components Jag2 Olfm4 Lgr5 Fzd2 Smoc2 Lfng Notch1 Hey1 Arid5b Gkn3 Tcf7 Igfbp4 Ascl2 Ephb3 Sox4 Tnfrsf19 Zfp503 c-Myc Rnf43 Ncor2 Jag1 Cdca7 Hes1 0.25 Fold change

ApcMin/+_/Ap4∆IEC _{vs. Apc}Min/+_/Ap4wt

ApcMin/+

Ap4wt Ap4ΔIEC

0 1 2 3

**

CT26 Ap4 IgG

**

*

**

*

_*

**

*

**

*

**

***

**

*

n.s.

Wnt/β-catenin components Notch components

% Input

d

**

IP Notch targets (Li et al. 2012) Lgr5+ stem cell signature

(Munoz et al. 2012) Wnt/β-catenin signaling (mSigDB) c-Myc targets (mSigDB) NES: –1.72; p < 0.001 NES: –2.32; p < 0.001 NES: –1.76; p = 0.001 NES: –1.61; p = 0.006 Up-regulated in Ap4ΔIEC Down-regulated in Ap4ΔIEC 0 2 4 6

c

Fold change (mRNA)

**

*

_***

*

***

**

*

**

ApcMin/+_/Ap4fl/fl

Ap4 EpCamSmoc2

Sox4 (A)Sox4 (B)Ascl2 (A)_{Ascl2 (B1)}_{Ascl2 (B2)}_{Ascl2 (C1)}_{Ascl2 (C2)}Tcf7 (A)Tcf7 (B)EphB3 Notch1 (A)Notch1 (B)

Dll1 (A) _{Jag1 (A)Jag1 (B)}_{Jag2 (A)} DII4 (A)DII4 (B)DII4 (C)

Cdkn1a (A1)Cdkn1a (A2)Cdkn1a (B)Cdkn1a (C) AchR Jag2 (B)Jag2 (C)

Hes1 Dll1 (B)

Lgr5Olfm4_Ctnnb1Sox4 Ascl2 Tcf7_EphB3Notch1Jag1 Jag2 Hey1c-Myc Hes1 p21 ApcMin/+/Ap4ΔIEC

Adenoma, Small intestine

n.s.

**

_**

**

Up-regulated

in Ap4ΔIEC Down-regulatedin Ap4ΔIEC

Up-regulated

in Ap4ΔIEC Down-regulatedin Ap4ΔIEC Up-regulatedin Ap4ΔIEC Down-regulatedin Ap4ΔIEC

Enrichment score (ES)

0.0 –0.1 –0.2 –0.3

Enrichment score (ES) Enrichment score (ES)

Ranked list metric (PreRanked) Ranked list metric (PreRanked)

–0.4 –0.5 –0.6 0.0 –0.1 –0.2 –0.3 –0.4 –0.5 –0.6 0.0 –0.1 –0.2 –0.3 –0.4 –0.5 2.5 –2.5 –5.0 0.0 2.5 –2.5 –5.0 0.0 0.0 0.1 –0.1 –0.2 –0.3 –0.4 –0.5 2.5 0.0 –5.0 0 2500 5000 7500

Rank in ordered dataset

10,000 12,500 15,000 0 2500 5000 7500

Rank in ordered dataset 10,000 12,500 15,000

0 2500 5000 7500

Rank in ordered dataset

10,000 12,500 15,000 0 2500 5000 7500

Rank in ordered dataset 10,000 12,500 15,000 –2.5 2.5 0.0 –5.0 –2.5 1 4

(7)

Paneth cells compared with ~5 Paneth cells in wild-type mice. As

determined by electron microscopy, Paneth cells of Ap4-deﬁcient

mice also displayed an increased number of vesicles, which

contain antimicrobial proteins, such as lysozyme and cryptdin

(Fig.

6 e). The length of the small intestine was increased in

Ap4-deﬁcient mice, presumably as a result of widened crypt bases

(Supplementary Fig. 5b). The length of the villi in the ileum was

slightly decreased when compared with wild-type mice

(Supple-mentary Fig. 5b). However, the number of TA cells in small

intestinal crypts (Supplementary Fig. 5b), the length of the colon

and the width of colonic crypts remained unchanged

(Supple-mentary Fig. 5c) The latter presumably due to the absence of

classical Paneth cells in the colon. Ap4 deﬁciency also resulted in

a decreased number of secretory goblet cells in villi of the small

intestine (Supplementary Fig. 5d) and in crypts of the small

intestine and colon (Supplementary Fig. 5e). Accordingly, mRNA

expression of stem cell markers (Smoc2, Lgr5, Olfm4) and the

goblet cell markers (Gob5, Muc2) was signiﬁcantly decreased,

whereas Paneth cell markers (Lysozyme, Cryptdin) were

sig-niﬁcantly increased in the epithelia of the small intestine of

Ap4-deﬁcient mice (Fig.

6 f). We did not detect any effect of Ap4

deletion on the rate of apoptosis or proliferation in the small

intestine (Supplementary Fig. 5f, g). Therefore, these processes

did presumably not cause the changes in the numbers of Paneth

cells, goblet cells and ISCs observed in Ap4-deﬁcient mice. In

addition, the effects of Ap4 deletion described here were

inde-pendent of the gender of the mice (Supplementary Fig. 5h, i).

Furthermore, IEC-speciﬁc deletion of Ap4 had the same effects on

the small intestinal and colonic architecture as the germ-line

deletion of Ap4 (Supplementary Fig. 6a-h). Therefore, the effects

of Ap4 loss on ISCs and their derivatives were intestinal epithelial

cell autonomous. Interestingly, Ap4 deﬁciency also decreased the

number of stem cells in the colon (Supplementary Fig. i, j).

Age-matched Apc

Min

mice deﬁcient for Ap4 also displayed a decreased

number of Lgr5- and Smoc2-positive ISCs per crypt, an increase

in Paneth cells, increased length of the small intestine and

enlargement of the crypt base of normal intestine, without any

change in proliferation or apoptosis in normal epithelium

(Sup-plementary Fig. 7a-g) independent of the gender (Sup(Sup-plementary

Fig. 7h). Notably, ISCs have been shown to efﬁciently form

intestinal tumors upon deletion of Apc

24

_{and play a critical role in}

adenoma and cancer cell self-renewal

25,26

_{. Taken together, these}

results suggest that the decreased rate of tumor formation in

Ap4-deﬁcient Apc

Min

_{mice is due to the lower number of functional}

ISCs in the intestinal crypts.

Analysis of

Ap4 function in intestinal organoids. To further

analyze the functional relevance of Ap4 for ISCs, we generated

small intestinal organoids by ex vivo culture of small intestinal

crypts derived from Villin-Cre-ERT2/Ap4

ﬂ/ﬂ

mice. After addition

of 4-OHT to established organoids, Ap4 expression was decreased

by ~90% within 3 days, which demonstrates efﬁcient,

Cre-mediated deletion of the

ﬂoxed Ap4 allele in these organoids

(Fig.

7 a). Upon acute Ap4 inactivation, organoids showed a

pronounced decrease of ISC markers, an increase of Paneth cell

markers, as well as a decrease of goblet cell markers within 3 days

after exposure to 4-OHT, whereas organoids derived from

Villin-Cre-ERT2/Ap4 wild-type mice exposed to 4-OHT did not display

signiﬁcant changes in the expression of these markers (Fig.

7 a).

Importantly, Ap4-deﬁcient organoids formed less protrusions

(crypt-like structures), when compared with the Ap4-expressing

organoids (Fig.

7 b). As the number of protrusions corresponds to

the number of self-renewing ISCs within organoids

27

_{, the}

decrease in the number of protrusions in Ap4-deﬁcient organoids

is presumably caused by a decrease in functional ISCs. Taken

together, these results suggest that Ap4 is essential for

main-taining ISCs in their undifferentiated state and plays an important

role in the homeostasis of ISCs and Paneth cells.

Gene expression proﬁling of Ap4-deﬁcient organoids. Next, we

obtained RNA expression proﬁles of Ap4-deﬁcient and Ap4

wild-type organoids 7 days after 4-OHT treatment using NGS.

Changes in gene expression observed after deletion of Ap4 were

considerably less pronounced in organoids when compared with

adenomas. By setting the cut-off for differential expression to a

fold change > 1.5 (p < 0.05), we identiﬁed 693 mRNAs as

differ-entially regulated as a consequence to deletion of Ap4 (Fig.

7 c),

with 319 mRNAs being signiﬁcantly downregulated, and

374 showing upregulation (Fig.

7 d). Remarkably, factors involved

in Notch signaling and Wnt/β-catenin signaling were signiﬁcantly

over-represented among the downregulated mRNAs

(Supple-mentary Fig. 8a, Supple(Supple-mentary Data 1). In line with the effect of

Ap4 deletion in adenomas, GSEA indicated that mRNAs

char-acteristic for Lgr5-positive ISCs and several factors involved in

Wnt/β-catenin and Notch signaling pathways were preferentially

downregulated upon deletion of Ap4 (Fig.

8 a, b, Supplementary

Fig. 8c, Supplementary Data 2): for example Sox4, Axin2, EphB3

were downregulated (Fig.

8 b, Supplementary Data 2).

Down-regulated components of the Notch signaling pathway included

Notch1, and the Notch target gene Hes1, as well as the Notch

activating ligands Dll1, Dll3, Dll4 and Jag2 (Fig.

8 b,

Supplemen-tary Data 2). Exemplary conﬁrmations of mRNA

down-regulations of genes important in Wnt/β-catenin signaling and/or

Notch signaling upon acute deletion of Ap4 in intestinal epithelial

cell derived organoids are shown in Fig.

8 c. The decreased activity

of the Notch pathway after Ap4 loss was conﬁrmed by

immu-nohistochemical detection of NICD1 in normal crypts in the

small intestine (Supplementary Fig. 8d, e). Not only was the

frequency of NICD1-positive cells per crypt lower, but also the

intensity of the NICD1 signal was decreased, which indicates a

lower activity of the Notch signaling pathway in these cells. Taken

together, these results show that Ap4 contributes to the

Wnt/β-catenin and Notch transcriptional program in normal intestinal

tissue.

Interestingly, the transcription factor Spdef (SAM pointed

domain-containing ETS factor) was downregulated in

Ap4-deﬁcient organoids according to NGS analysis and validated by

qPCR (Fig.

8 b, c). Spdef regulates the differentiation and

Fig. 4 Ap4-dependent expression profiles in ApcMin/+adenomas.a GSEA comparing gene expression profiles from ApcMin/+/Ap4fl/fland ApcMin/ +_/Ap4ΔIEC_{adenomas from 120 days old mice with Lgr5-positive stem cell signatures}13_{, Wnt/}_{β-catenin signaling (mSigDB: molecular Signatures Database),}

Notch target genes or c-Myc target genes (mSigDB). NES: normalized enrichment score, Nom. p-value: nominal p-value.b Heatmap of selected differentially expressed mRNAs (p-value < 0.05) from intestinal stem cell gene signatures, Wnt/β-catenin signaling and/or Notch signaling gene signatures analyzed ina. The heatmap displays relative fold changes in expression levels normalized to the mean expression in the control, ApcMin/ +_/Ap4ﬂ/ﬂ_{, samples for each indicated mRNA. Three biological replicates per genotype were analyzed.}_{c qPCR analysis of the indicated mRNA derived from}

tumors from three female mice (five tumors per mouse) per genotype. d The murine CRC cells CT26 were subjected to qChIP analysis with Ap4 or IgG-specific antibodies for ChIP. The mouse acetylcholine receptor (AchR) promoter, which lacks Ap4-binding motifs, served as a negative control. E-boxes used for qChIP analysis are marked in Supplementary Fig. 3.c, d Results represent the mean ± SD. Results were subjected to an unpaired, two-tailed Student’s t-test with p-values * < 0.05, ** < 0.01, *** < 0.001, n.s.: not significant. See also Supplementary Fig. 3, Supplementary Data 1 and Supplementary Data 2

(8)

maturation of goblet cells

28

. Downregulation of Spdef may

therefore contribute to the decreased number of goblet cells

observed in Ap4-deﬁcient mice.

As observed in adenomas, c-Myc and E2F target genes were

signiﬁcantly enriched among the downregulated RNAs in

AP4-deﬁcient organoids (Fig.

8 a, Supplementary Fig. 8c,

Supplemen-tary Data 2) albeit with rather modest fold changes in expression

that were considerably less pronounced compared with those of

ISC signature and Notch target genes (Supplementary Fig. 8f,

Supplementary Data 2). Interestingly, the differential mRNA

expression caused by deletion of Ap4 was similar in organoids

and adenomas as determined by correlation analysis (Fig.

8 d). As

the organoid derived expression proﬁles were obtained in the

absence of non-epithelial cells or stroma, these

ﬁndings indicate

that the gene expression changes resulting from the inactivation

of Ap4 are largely epithelial cell autonomous. These results

suggest that the differential regulation of factors involved in Wnt/

β-catenin and/or Notch signaling observed after deletion of Ap4

in Apc

Min

_{mice occurs in normal intestinal epithelial stem cells}

prior to adenoma development.

c

a

0 10 20 30 Tu m o ro id s / 50 μ l m a trigel

***

0 2 4 n.s .

**

_**

*

_**

_***

_**

*

_**

**

d

F o ld c hange (mRNA) Tumoroids Pas s age 0 Lgr 5 0 10 20 30 40

*

Lgr 5 pos . area (%) ApcMin/+ ApcMin/+ ApcMin/+

ApcMin/+_/Ap4fl/fl _ApcMin/+_/Ap4ΔIEC

ApcMin/+/Ap4fl/fl ApcMin/+/Ap4ΔIEC

0 5 10 15 20 n.s. Tu m o ro id s / 50 μ l m atr igel Pas s age 1

g

f

0 200 400 600 800 1000 M e an siz e (μ m) n.s. n.s. n.s. n.s. n.s.

h

Pas s age 1 P a ssa g e 5

e

55 72 17 36 28 72 55

Ap4wt_Ap4IEC

Mouse 250 1 2 3 4 1 2 3 130 95 - Nicd1 - Ap4 - Hes1 -α-Tubulin (KD)

b

Fold c h ange (mRNA) 0 1

2 ApcMin/+_/Ap4fl/fl

ApcMin/+_/Ap4IEC

ApcMin/+_/Ap4fl/fl

ApcMin/+/Ap4ΔIEC

Tumoroids

***

n.s .

**

0 10 20 30 40 50 1 2 3 4 5 Passage: 1 2 3 4 5 Passage: Tu m o ro id s / 25 μ l mat rigel

***

n.s. n.s. n.s. n.s. Apc–/–_/Ap4+/+ Apc–/–_/Ap4–/– Apc–/–_/Ap4+/+ Apc–/–_/Ap4–/–

Apc–/–_/Ap4+/+ _Apc–/–_/Ap4–/–

Ap4 fl/fl

Ap4

Ctnnb1 Sox4 AscI2 Tcf7 EphB3 Notch1 Jag1 Jag2 Hey1 c-Myc Hes1 p21

EpCam Smoc2 Lgr5 Olfm4

Ap4 ΔIEC Ap4 fl/fl Ap4 ΔIEC

(9)

Regulation of the NOTCH pathway by AP4 in human CRC

cells. In order to determine whether the connection between Ap4

and Notch detected here is conserved between species and

rele-vant to human CRCs, we performed expression and functional

analyses in human CRC cell lines. In line with the results

described above, the expression of AP4 and NICD1 proteins

positively correlated in a panel of

ﬁve CRC cell lines

(Supple-mentary Fig. 9a). After ectopic AP4 expression in DLD-1 cells,

NOTCH1/NICD1 and the NOTCH-target genes HES1 and

NRARP were induced at the protein and mRNA levels

(Supple-mentary Fig. 9b, c). Downregulation of AP4 by RNA interference

decreased NICD1 protein expression in Colo320 cells

(Supple-mentary Fig. 9d). Furthermore, a NOTCH activity reporter

plasmid was induced by ectopic AP4 expression, but not by a

mutant AP4 protein, which lacks the basic DNA-binding region

(Supplementary Fig. 9e). In addition, analysis of public ChIP-Seq

data showed open and active chromatin surrounding the sites of

AP4 occupancy at the NOTCH1 promoter, because histone

H3K4me1 and H3K27Ac modiﬁcations were increased in their

vicinity (Supplementary Fig. 9f). When we analyzed ChIP-Seq

data, which we had previously obtained after ectopic AP4

expression in the CRC cell line DLD-1

7

_{, we detected AP4}

occu-pancy at the ASCL2, DLL1, DLL4, EPHB3, HES1, JAG1, JAG2,

NOTCH1, SOX4 and TCF7 promoters in human DLD-1 CRC

cells (Supplementary Fig. 9g). Therefore, AP4 regulates genes

involved in WNT/β-catenin and/or NOTCH signaling directly by

binding to their promoters in CRC cells.

In addition, inhibition of NOTCH signaling by exposure to the

γ-secretase inhibitor Dibenzazepine (DBZ) resulted in a decrease

of AP4 and c-MYC expression in SW620 CRC cells

(Supplemen-tary Fig. 9h). As expected, the NOTCH-target genes NRARP and

HES1 were also repressed. Notably, DBZ suppressed Ap4 protein

expression in the small and large intestine of mice

(Supplemen-tary Fig. 9i). As expected, inhibition of Notch signaling led to an

increase in the number of Paneth cells and goblet cells, as well as

to a decrease of Hes1 and c-Myc expression. As we could not

obtain experimental evidence for a direct regulation of AP4 by

NICD1 (data not shown), the regulation of AP4 by the NOTCH

pathway is presumably mediated by c-MYC, which represents a

known target of the NOTCH pathway

29,30

. Therefore, AP4, the

NOTCH pathway and c-MYC form a positive feed-back loop.

Role of AP4 in human CRCs. To obtain further evidence for a

clinical relevance of AP4 in CRC initiation and progression, we

analyzed patient-derived expression data that were generated by

the TCGA consortium

31

_{. Indeed, AP4 mRNA expression was}

signiﬁcantly increased in primary CRCs when compared with

normal mucosa in 41 matched normal versus CRC patient

samples, as well as in unmatched patient samples representing

normal mucosa (n

= 41) and primary CRCs (n = 462) (Fig.

9 a).

Recently, CRCs were shown to belong to four different molecular

subgroups, the so-called consensus molecular subtypes 1–4

32

_{. In}

line with the results obtained here, CRCs belonging to the

CMS2 subtype, which is characterized by high WNT and c-MYC

activity, showed signiﬁcantly elevated expression of AP4 when

compared with the three other CMS subtypes (Supplementary

Fig. 10a). Moreover, in CRCs from the TCGA cohort AP4

expression showed a positive correlation with the expression of

mRNAs characteristic for Lgr5-positive ISCs

14

, as well as with

mRNAs encoding factors involved in Wnt/β-catenin signaling

and c-MYC target genes (mSigDB, molecular signature

data-base:

33

; Fig.

9 b). Furthermore, AP4 expression showed a

sig-niﬁcant positive correlation with ASCL2, TCF7, NOTCH1 and

JAG2 expression in 462 primary CRCs in the TCGA cohort

(Fig.

9 c). As expected, AP4 expression was also positively

asso-ciated with c-MYC and negatively assoasso-ciated with CDKN1A/p21

expression.

In order to determine whether the positive correlation between

AP4 and NOTCH1/NICD1/HES1 also exists on the level of

protein expression in human CRCs, we determined AP4, NICD1

and HES1 expression levels by immunohistochemical analysis of

220 primary CRC samples. For the evaluation of AP4, NICD1

and HES1 expression, we established a four-stage scoring scheme

(Supplementary Fig. 10b). AP4 expression was highly concordant

with both NICD1 and HES1 expression (Fig.

9 d) indicating that

the reciprocal regulation between AP4 and the NOTCH pathway

also occurs in primary, human CRCs.

Discussion

This study identiﬁed Ap4 as an important, rate-limiting mediator

of intestinal adenoma initiation (summarizing model in Fig.

10 ).

As inactivation of Ap4 led to a decrease in the number of ISCs

and an increase in Paneth cells, it is conceivable that the

decreased formation of adenomas in the absence of Ap4 is due to

the smaller number of bona-ﬁde ISCs that are able to initiate

adenomas after acquiring further genetic and epigenetic

altera-tions. This hypothesis is in accordance with the observation that

tumor-promoting mutations in ISCs are more efﬁcient in

gen-erating tumors than mutations in other, more differentiated

intestinal epithelial cells

24,34

_{. In addition, recent analyses have}

provided further support for a direct role of stem cell abundance

in the determination of tumor frequencies

35

: that is, a strong

correlation was found between the tissue-speciﬁc stem cell

number and the risk to develop a tumor for this particular tissue.

Unexpectedly, deletion of Ap4 in intestinal epithelial cells had

no effect on cellular proliferation. In contrast, acute deletion of

Fig. 5 Deletion of Ap4 decreases stemness in adenomas and tumor organoids. a Left panel: in situ hybridization of Lgr5 mRNA. Scale bars represent 100µm. Right panel: quantification of Lgr5-positive area in % in the adenomas from two male and one female mice in at least six adenomas per genotype. b Left panel: representative pictures of small intestinal tumor organoids 6 days after isolation (passage 0), upper panel, or 4 days after passaging (passage 1), lower panel. Organoids were isolated from three tumors per mouse from two female and two male mice per genotype. Scale bars represent 500_µm. Right panels: number of tumor organoids per drop of 50µl Matrigel. A total of 24 drops (passage 0) or 15 drops (passage 1) was analyzed per genotype. c, d qPCR analysis of the indicated mRNA derived from tumor organoids. e Western blot analysis of the indicated proteins. f Representative pictures of tumor organoids derived from small intestinal epithelial cells obtained from Lgr5-CreERT2+/-/Apcfl/fland Lgr5-CreERT2+/-/Apcfl/fl/Ap4fl/flmice after treatment with 4-OHT. After isolation, organoids were kept in ENR media (contains EGF, Noggin and RSPO1). Forty-eight hours after isolation, organoids were treated with 4-OHT in a concentration of 100 nM for 48 h to delete Apc or Apc in addition to Ap4 in Lgr5-positive intestinal stem cells (ISC). Additional 48 h later, organoids were passaged and EN media (containing EGF and Noggin, but without RSPO1) was used, which selectively allowed Apc-deficient tumoroids to expand (passage 1). Pictures were taken 7 days after passaging in case of passage 1, and 6 days after passaging in case of passage 5. g Mean tumor organoid number per drop of 25µl Matrigel calculated as a mean of a total of 20 drops of 25 µl Matrigel each for passage 1, 3–5 or a total of 11 drops (Ap4 wt) and 6 drops (Ap4 ko) for passage 2.h Mean tumor organoid size was measured and calculated from Matrigel drops as depicted exemplarily ind. a, b, c, d, g, h Results represent the mean ± SD. Results were subjected to an unpaired, two-tailed Student’s t-test with p-values * < 0.05, ** < 0.01, *** < 0.001, n.s.: not significant. See also Supplementary Fig. 4

(10)

0 2 4 6 8 10

***

Paneth cells/crypt

d

a

b

c

f

e

0 10 20 30 Vesicle/Paneth cell

***

Ap4 Small intestine 0 2 4 6

*

**

*

n.s.

***

**

Ap4+/+ _IEC Ap4–/– _IEC

Small intestinal IEC

Fold change (mRNA)

0 2 4 6 8 LGR5 -GFP pos. cells/crypt

***

Lgr5 -GFP Small intestine +/+ –/– +/+ –/– Ap4 +/+ –/– Ap4 +/+ +/– –/– Ap4 Lysozyme Ap4–/– Ap4–/– Ap4+/+ Ap4+/+ Ap4–/– Ap4+/+ Ap4–/–

Ap4+/+ _Ap4+/+ _Ap4–/–

Ap4 0 2 4 6 8 Olfm4 Olfm4 pos. cells/crypt *** Small intestine Ap4

EpCam Smoc2 Lgr5 Olfm4

LysozymeCryptdin

Gob5 Muc2

Fig. 6 Inactivation of Ap4 causes decrease of ISC and increase Paneth cell numbers. a Immunohistochemical detection of Ap4 (brown) in small intestinal tissue, ileum of one male and one female mouse per genotype. Scale bar= 50 µm (25 µm insert), white arrow: site of specific Ap4 expression. Mast cells in the villi display an unspecific staining. Counterstaining with hematoxylin. b Left panel: immunohistochemical analyses of Lgr5-eGFP in intestinal sections of 63 days old Lgr5-eGFP mice. Scale bars represent 25µm. Right panel: quantification of Lgr5-eGFP-positive cells in the crypt base of the ileum from two male and two female mice (130 crypts) per genotype.c Left panel: in situ hybridization of Olfm4 mRNA. Scale bars represent 25µm. Right panel: quantification of Olfm4-positive cells in the crypt base from two male and two female mice (316 crypts) per genotype. d Left: immunohistochemical detection of Lysozyme (brown) expressed in Paneth cells. Counterstaining with hematoxylin. Scale bar: 50_{µm (25 µm insert). Right: the small intestine/} ileum from two male and two female mice (100 crypts) per genotype was analyzed for Paneth cells/crypt.e Electron microscopic analysis of small intestinal crypt base; white arrow: Paneth cells. Scale bar: 25µm. f qPCR analysis of the indicated mRNAs in IECs of the ileum of two male and one female mice per genotype.b, c, d, e, f Results represent the mean ± SD. Results were subjected to an unpaired, two-tailed Student_{’s t-test with p-values * < 0.05,} ** < 0.01, *** < 0.001, n.s.: not significant. See also Supplementary Figs. 5, 6, 7

(11)

c-Myc using Ah-Cre or Villin-CreER alleles in post-natal IECs

resulted in defects in proliferation and biosynthetic activity in the

small intestine

36,37

_{. The different result may be due to the}

dif-ferent timing of gene inactivation: here we used either germ-line

or Villin-Cre-mediated deletion of Ap4, whereas the two studies

on c-Myc employed deletion at least 7 days after birth and later,

which was necessary because c-Myc is essential during

embry-ogenesis. However, they observed that IEC proliferation was at

least in part independent of c-Myc. Therefore, IECs may rely on

alternative pathways for promoting cell proliferation. In addition,

these differences indicate that Ap4 is responsible for mediating a

distinct aspect of c-Myc function and does not simply perform all

functions of c-Myc.

The results obtained in tumoroids derived from Apc

Min

_mice

and in tumoroids generated by acute deletion of Apc imply that

Ap4 is important for the initiation but not required for the

maintenance of tumoroids. These results are in line with the

observations we made in Apc

Min

_{mice, where Ap4 loss decreased}

b

Vil-CreERT2 Vil-CreERT2/Ap4fl/fl + 4-OHT 0 5 10 15 20 Protrusions/organoid

***

d

a

c

693 mRNAs DESeq2 (735 mRNAs) edgeR (1047 mRNAs) Fold change ≥1.5× p < 0.05

log2 fold change Ap4–/– vs. Ap4fl/fl

–log 10 (p -v alue) 0 2 4 6 –4 –2 0 2 4 8 mRNAs Min Max Rel. expression Ap4–/– Ap4fl/fl 1 2 3 1 2 3 n = 374 n = 374 n = 319 n = 319 0 2 Ap4 4 6

*

**

*

***

Vil-CreERT2/Ap4fl/fl n.s.

Fold change (mRNA)

*

_**

n.s.

_*

**

* *

*

n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. 3 days 7 days 0 days Vil-CreERT2 4-OHT:

*

EpCam Smoc2 Lgr5 Olfm4_Cryptdin Gob5 Muc2 Ap4EpCam

Vil-CreERT2 Vil-CreERT2/Ap4

fl/fl

Smoc2 Lgr5 Olfm4_Cryptdin Gob5 Muc2

Fig. 7 Effects of Ap4 deletion on intestinal organoids. a qPCR analysis of the indicated mRNA from organoids isolated from three female mice per genotype after passaging and 7 days after Cre activation by 4-OHT.b Left panel: representative pictures of small intestinal organoids 7 days after passaging and Cre activation by 4-OHT. Scale bars represent 20µm. Right panel: quantification of protrusions (crypt-like structures) per organoid derived from three female mice per genotype. A total of 81 organoids were evaluated per genotype.c Venn diagram displaying differentially regulated mRNAs (fold change_{≥ 1.5,} p < 0.05) in Ap4fl/fland Ap4ΔIECorganoids as determined by edgeR and DESeq2.d Volcano plot and heatmap depicting expression changes of differentially expressed mRNAs between Vil-Cre-ERT2 and Vil-Cre-ERT2/Ap4fl/florganoids, 7 days after Cre activation by 4-OHT, detected by RNA-Seq. Left panel: volcano plot depicting expression changes of differentially expressed mRNAs (fold change_{≥ 1.5) from Ap4}fl/fland Ap4ΔIECorganoids. Downregulated mRNAs are depicted in blue, upregulated mRNAs are depicted in red. RNAs with fold changes < 1.5 and/or statistically nonsignificant changes in expression are indicated in black. Dashed vertical lines indicate 1.5-fold change cutoff. Dashed horizontal line indicates the cutoff for adjusted p-values < 0.05 as determined with DESeq2. Right panel: Heatmap depicting expression changes of differentially expressed mRNAs (fold change_{≥ 1.5 p < 0.05 as} determined by edgeR and DESeq2) from Ap4fl/fland Ap4ΔIECorganoids. Colors indicate relative expression values from minimum (blue) to maximum (red) for each RNA sample per differentially regulated mRNA.a, b Results represent the mean ± SD. Results were subjected to an unpaired, two-tailed Student’s t-test with p-values * < 0.05, ** < 0.01, *** < 0.001, n.s.: not significant

(12)

the number of adenomas but not their growth/size. Interestingly,

Lgr5 gene expression has been reported previously to be

dis-pensable for ex vivo tumor organoid maintenance

38

_.

Notably, the deletion of Ap4 resulted in decreased expression of

Notch pathway components including Notch1 itself.

Interest-ingly, Notch pathway inactivation promotes differentiation of

ISCs into Paneth cells

21

, leading to Paneth cell hyperplasia and a

decrease in ISCs

21,39

. However, the number of goblet cells

increases after Notch inhibition, whereas it decreased after Ap4

inactivation in our study. Therefore, Ap4 loss does not simply

recapitulate inhibition of the Notch pathway suggesting that Ap4

regulates additional pathways/genes, which contribute to the

differentiation of goblet cells. For example, Ap4 deﬁciency

resulted in a decrease in expression of the transcription factor

Spdef. Interestingly, it was previously shown that ectopic

expression of Spdef in the murine intestine promotes the

a

b

c

Stem cell signature Wnt/β-cat. components Notch components Fold change

Ap4ΔIEC vs. Ap4wt

0.25 1 2 3 Ap4fl/fl _Ap4–/– Olfm4 Fzd2 Dll3 Spdef Rgmb Gkn3 Smoc2 Jun Notch1 Axin2 Lgr5 Dll1 Ascl2 Lfng Notch3 Sox4 Dll4 Notch4 Cdca7 Zfp503 Lrig1 Ephb3 Dtx3 Igfbp4 Rnf43 Sox9 Hes1 Dtx4 Jag2 1 2 3

d

log

2_{fold change organoid}

Ap4

–/–

vs

Ap4

fl/fl

log₂fold change tumor

Ap4ΔIEC vs Ap4fl/fl_(ApcMin/+₎

0 –0.5 –1 –1.5 –2 0 –0.5 –1 –1.5 –2

Lgr5+/EphB2high stem cell signature (Merloz-Suarez et al. 2011) Notch target genes (Li et al. 2012)

c-Myc target genes (MSigDB) Pearson r = 0.523

P < 0.0001

NES: –2.70; p < 0.001 NES: –1.79; p = 0.002 Lgr5+ stem cell signature

(Munoz et al. 2012) c-Myc targets (mSigDB) NES: –1.53; p = 0.014 Wnt/β-cat./ISC markers (Fevr et al. 2007) NES: –2.17; p < 0.001 Direct Notch targets

(Li et al. 2012)

0 1 2

*

Vil-CreERT2/Ap4fl/fl 3 days 0 days

Fold change (mRNA)

4-OHT:

* *

*

_**

* *

_*

_{* *}

*

**

n.s. n.s. Up-regulated in Ap4ΔIEC Down-regulated in Ap4ΔIEC Up-regulated in Ap4ΔIEC Down-regulated in Ap4ΔIEC Up-regulated in Ap4ΔIEC Down-regulated in Ap4ΔIEC Up-regulated

in Ap4ΔIEC Down-regulatedin Ap4ΔIEC 0.0

–0.1

Enrichment score (ES) Enrichment score (ES)

Ranked list metric (PreRanked)

Ranked list metric (PreRanked) Ranked list metric (PreRanked)

Ranked list metric (PreRanked)

–0.2 –0.3 –0.4 –0.5 –0.6 2 1 –1 0 2500 5000 7500 10,000 12,500

Rank in ordered dataset Rank in ordered dataset

0 2500 5000 7500 10,000 12,500 0 2500 5000 7500 10,000 12,500 0 2500 5000 7500 10,000 12,500 –2 0 2 1 –1 –2 0 2 1 –1 –2 0 2 1 –1 –2 0 0.0 –0.1 –0.2 –0.3 –0.4 –0.5 –0.6 0.0 0.10 0.05 0.00 –0.05 –0.10 –0.15 –0.20 –0.25 –0.30 –0.35 –0.40 –0.45 –0.1 –0.2 –0.3 –0.4 –0.5 –0.6 Ctnnb1 So x4

Axin2 Ascl2 _Notch1 Dll1 Jag2 Hes1 Dll4 Spdef

4 1

(13)

differentiation of goblet cells, whereas it decreased the number of

Paneth cells

28

. Accordingly, the decrease of Spdef observed in

Ap4-deﬁcient organoids could potentially also contribute to the

decreased number of goblet cells detected after Ap4 deletion.

Alternatively, the decrease in goblet cells may have been a

com-pensatory response to the increase in Paneth cells caused by Ap4

deletion.

As activation of Notch signaling promotes the initiation of

murine intestinal adenomas

40

_{and inhibition of Notch signaling}

leads to mitotic arrest and apoptosis in human colon cancer

cells

41

, the Notch pathway may represent a route via which Ap4

promotes intestinal adenoma initiation. Conversely, Ap4 deletion

may prevent adenoma initiation via inhibiting the Notch pathway.

By studying human CRC cell lines, we had previously found

that ectopic Ap4 expression is sufﬁcient to activate the Wnt

pathway. Our NGS analysis presented here further supports these

ﬁndings with in vivo evidence, because numerous components of

the Wnt/β-catenin signaling pathway were downregulated in

Ap4-deﬁcient adenomas and derived organoids. In support of this

conjecture, a comprehensive study has recently shown that Ap4 is

an important component of Wnt signaling during X. laevis

development and acts down-stream of the

β-catenin destruction

complex to regulate expression of Wnt/β-catenin target genes

42

_.

Therefore, Ap4 presumably represents an integral component of

the Wnt pathway and its activities during development and

tumorigenesis. The downregulation of genes involved in Wnt

signaling observed here may critically contribute to the decreased

number of ISCs observed in Ap4-deﬁcient mice, because the Wnt

pathway has been implicated in the maintenance of stemness and

suppression of differentiation in ISCs

43–45

.

It is well known that a controlled balance of the Notch and

Wnt pathway activity is important for the homeostasis of stem

cells and cell fate decisions in the intestine and that these two

pathways regulate each other at multiple points

46

. Speciﬁcally,

inhibition of NOTCH1/2 receptors was shown to induce Wnt

signaling, which then promoted goblet cell differentiation

46

. The

decrease of Wnt pathway component expression after AP4

dele-tion may therefore alter the expected outcome of a Notch

inhi-bition (i.e., increase of all secretory cell numbers) unto the

decrease of goblet cell differentiation observed here.

The increased number of Paneth cells in Ap4-deﬁcient mice

might result from an increased propensity of Ap4-deﬁcient ISCs

to generate Paneth cell precursors during asymmetric stem cell

division. Interestingly, Lgr5 represents an Ap4 target gene

7

and

Lgr5 deﬁciency promotes Paneth cell differentiation in mice

47

_.

Further discussions of the results and the potential limitations

of the Apc

Min

mouse model can be found in the Supplementary

Discussion.

In conclusion, our study demonstrates an unexpected, central

role of Ap4 in ISCs and Paneth cell homeostasis and revealed that

Ap4 function is critical for adenoma initiation in a preclinical

model of inherited colon cancer. Besides illuminating an

impor-tant aspect of CRC biology, our results indicate that Ap4

repre-sents a candidate therapeutic target for the treatment of CRCs.

Methods

Generation and husbandry of mice. Targeted ES cells with C57BL/6N back-ground were obtained by homologous recombination with a vector containing the Ap4 exons 2–4 flanked by loxP sites and an intronic neomycin resistance (Neo) cassetteflanked by frt sites (scheme in Jackstadt et al.6_{). Ap4}fl/fl_{mice were}

gen-erated by injection of targeted ES cells into C57BL/6N blastocyst. The Neo cassette was removed by crossing toﬂp-mice48_{and germ-line Ap4 knock-out mice were}

generated by crossing with CMV-Cre+/− mice49_{. Ap4}-/-_{mice showed no overt} phenotype and were born at normal Mendelian ratio. Oligonucleotides used for genotyping are listed in Supplementary Table S1. For analysis of the effect of Ap4 inactivation on the ISC number, we used Lgr5-eGFP-Cre-ERT2+/− mice50

(obtained from Hans Clevers, University Medical Center Utrecht, The Nether-lands) and for speciﬁc deletion of Ap4 in intestinal epithelial cells or derived organoids, we used Villin-Cre+/− or Villin-Cre-ERT2+/− mice (obtained from Klaus-Peter Janssen, Technical University Munich, Germany), respectively51_.

APCMin/+mice9,10_{were used to analyze the role of Ap4 in intestinal adenoma}

development (obtained from Marlon Schneider, Ludwig-Maximilians-Universität, München, Germany). Mice were kept in individually ventilated cages with a 12-h light/dark cycle and ad libitum access to water and standard rodent diet. For determination of proliferation rates, 75 mg/kg BrdU (Amersham) in phosphate-buffered saline (PBS) was intraperitoneally injected 1.5 h before mice were sacri-ﬁced. All animal experimentations and analyses were approved by the Government of Upper Bavaria, Germany (AZ 55.2-1-54-2532-4-2014).

Tissue preparation and adenoma counting. After isolation of intestinal tissue, the colon and small intestine were separated andflushed with PBS to remove stool. The small intestine was dissected into duodenum, jejenum and ileum. The colon and small intestine were opened longitudinally and rolled with the mucosa oriented outwards andfixed in formalin, dehydrated and embedded into paraffin. For evaluation of tumor numbers, each part of the intestine was cut longitudinally and spread on Whatman 3 MM paper. Afterfixation in formalin, adenomas were counted under a dissection microscope (Zeiss) with 10× magnification. Hematoxylin and eosin (HE) and PAS/Alcian blue staining. Formalin-fixed, paraffin-embedded (FFPE) tissue was cut into 2 µm sections on a rotating microtome (Microm HM355S, Thermo Scientific). The slides were de-paraffinized and stained with hematoxylin (Waldeck) for 6 min followed by eosin (Sigma-Aldrich) for 2.5 min in an automated slide staining device (Tissue-Tek, Prisma). Periodic acid-Schiff (PAS) staining was done by applying Alcian blue pH 1 (Bio Optica) for 10 min followed by periodic acid (Merck) for 5 min, Schiff’s reagent (Sigma-Aldrich) for 5 min and counterstaining with hematoxylin (Waldeck).

Immunohistochemistry. FFPE tissue was cut into 2 µm sections on a microtome and de-parafﬁnized. After antigen retrieval, slides were incubated with primary antibody (the primary antibodies used are listed in Supplementary Table 4) for 1 h at room temperature and washed with TRIS-HCL buffer (pH 7.5) followed by a secondary antibody. Antibodies were detected with the ABC kit using DAB (Vector and Dako) for brown stainings or AEC (Thermo Fisher Scientiﬁc) for red stainings. The slides were counterstained with hematoxylin (Vector) and mounted with Roti®-Histokitt II (Roth). All stainings were performed with the respective IgG control (Supplementary Table 4) as a negative control and without primary antibody as a system control. Images were captured on an Axioplan2 imaging microscope (Zeiss) equipped with an AxioCamHRc Camera (Zeiss). For analysis of cleaved caspase-3, the AxioVision Software (Zeiss) was used to measure the area for each tumor in mm2_.

Fig. 8 NGS analysis of intestinal organoids after deletion of Ap4. a GSEA comparing gene expression profiles from Vil-Cre-ERT2 and Vil-Cre-ERT2/Ap4fl/fl organoids 7 days after Cre activation by 4-OHT with Lgr5-positive stem cell signatures13_,_{β-catenin regulated/ISC-specific genes}60_{, Notch targets genes}61

or c-Myc target genes (mSigDB: molecular Signatures Database). NES: normalized enrichment score, Nom. p-value: nominal p-value.b Heatmap depicting expression changes of selected differentially expressed mRNAs (p-value < 0.05) from stem cell gene signatures, Wnt signaling and/or Notch signaling gene signatures analyzed ina. The heatmap displays relative expression levels normalized to the mean expression in the control (Vil-CreERT2) samples for each mRNA. Three biological replicates per genotype were analyzed.c qPCR analysis of the indicated mRNA of organoids with the indicated genotypes 3 days and 7 days after 4-OHT induction.d Scatter plot displaying the correlation of expression changes of the 424 mRNAs significantly (p < 0.05) downregulated in both adenomas and organoids (shown in gray).The mRNAs from stem cell gene signatures, Notch target gene signatures and c-Myc target genes analyzed in Figs.4a and8a in both adenomas and organoids are highlighted with the indicated colors. The Pearson correlation coefficient of all 424 mRNAs downregulated in both adenomas and organoids and statistical significance are indicated. c Results represent the mean ± SD. Results were subjected to an unpaired, two-tailed Student’s t-test with p-values * < 0.05, ** < 0.01, *** < 0.001, n.s.: not significant. See also Supplementary Fig. 9, Supplementary Data 1 and Supplementary Data 2